A method for designing brewing formula of tangerine red wine based on dynamic optimization algorithm

Abstract

The invention discloses a method for designing a formula for brewing jujube red wine based on a dynamic optimization algorithm, comprising the following steps: S1, constructing an initial multidimensional data set; S2, generating an initial formula setting by a fuzzy adaptive co-evolution algorithm; S3, dividing the brewing process into an initial fermentation, a mid-term fermentation, and a late maturation stage; S4, using an improved particle swarm optimization algorithm to control the exploration and development balance of each particle, and using a multi-agent collaborative mechanism for real-time optimization; S5, dynamically adjusting the variation amplitude by applying an adaptive variation distribution mechanism at each stage; S6, optimizing specific brewing targets at different brewing stages by a staged multi-objective optimization system; S7, applying a traceability analysis system to analyze each batch of brewing data; S8, adjusting the initial formula setting of the next batch based on the traceability analysis results. The invention uses a fuzzy adaptive co-evolution algorithm and an improved particle swarm optimization algorithm, etc., to achieve multi-stage intelligent control of the brewing process of jujube red wine.

Description

Technical Field

[0001] The invention relates to the technical field of brewing process optimization, and in particular to a method for designing a brewing formula of tangerine red wine based on a dynamic optimization algorithm. Background Art

[0002] In the brewing process of fermented wines such as Huajuhong wine, the differences in raw material ingredients, changes in the brewing environment, and the complexity of fermentation conditions all have an important impact on the quality of the finished product. Traditional brewing methods mostly rely on the experience of the winemaker, combining sensory judgment and some physical and chemical tests to determine the brewing conditions. Although this method has accumulated certain experience in practice, due to the complexity of the process and the diversity of brewing factors, it is difficult to achieve high-precision and stable control by relying solely on experience. Therefore, the traditional brewing process has shortcomings in dealing with the consistency and standardization of the quality of different batches of wines. Especially in the face of the market demand for high-quality, stable-tasting Huajuhong wine, the limitations of the existing process become particularly obvious.

[0003] In order to solve these problems, some studies and processes in recent years have tried to introduce mathematical models and computer control technology into the brewing process. For example, quality control technology based on multidimensional data analysis and applied statistical methods have achieved certain results in some brewing links. These methods monitor brewing environmental parameters such as temperature, humidity, pH value, etc. during the fermentation process, analyze the dynamic changes of the fermentation process, and try to establish a simple quality prediction model based on this. However, since most traditional methods remain at the level of simple correlation analysis and local optimization, they fail to achieve the coordinated optimization of multiple brewing stages and multiple parameters, making it difficult to effectively control the entire brewing process. In addition, with the gradual complexity of brewing technology and the improvement of wine quality requirements, the limitations of existing models have become more and more prominent, especially in the dynamic adjustment and adaptation of multi-objective optimization in different brewing stages, which has not yet achieved ideal results.

[0004] Traditional optimization methods such as genetic algorithms and particle swarm algorithms have been applied to the automatic control of some brewing processes to adjust a single or a few key parameters and optimize the objectives of a fixed stage. These methods have improved the brewing efficiency to some extent and have produced certain effects on the parameter control of a single stage. However, genetic algorithms and particle swarm algorithms have shown obvious shortcomings when facing the multi-stage dynamic, nonlinear and multi-objective optimization requirements of the brewing process. On the one hand, traditional genetic algorithms are difficult to cope with the rapidly changing requirements during the fermentation process and have limited parameter adaptability; on the other hand, particle swarm algorithms uniformly control parameters in different stages and fail to perform targeted optimization control for specific requirements such as sugar conversion in the early stage of fermentation, flavor generation in the middle stage, and texture balance in the later stage. The existing single algorithm is difficult to meet the multi-objective optimization requirements of multiple brewing stages, and lacks an intelligent and adaptive control mechanism to dynamically adjust brewing parameters.

[0005] In addition, in the field of brewing data analysis, the existing traceability analysis system also has certain shortcomings. Traceability analysis has been applied in the food, beverage and other industries, but in the brewing process of fermented wines, especially in brewing varieties such as orange red wine that have extremely high taste requirements, traditional traceability analysis systems can only record simple parameter fluctuations and lack effective tracking and analysis capabilities for subtle differences between different batches. In dealing with the uncertainty of multiple parameters and the brewing process, traditional traceability analysis is mostly limited to the backtracking of static data, and lacks real-time analysis and adaptive control capabilities for dynamically changing data during the fermentation process. This has led to the traceability analysis system failing to effectively discover the key parameters that affect product consistency, and failing to establish a multi-dimensional, multi-level data association analysis model in the brewing process.

[0006] As the demand for Huajuhong wine in the market gradually increases, consumers' requirements for the consistency of its taste, flavor, sweet-sour ratio and texture are also increasing. Existing technologies have obvious defects in meeting these requirements: on the one hand, since traditional brewing relies on manual regulation and experience accumulation, it is difficult to meet the consistency requirements of multiple batches of products; on the other hand, the existing automated control methods are not adaptable enough in complex fermentation processes, especially when facing the needs of collaborative optimization of multiple brewing stages and multiple parameters, traditional optimization algorithms lack pertinence. At the same time, since the brewing process of Huajuhong wine is affected by differences in raw material ingredients, environmental changes and fluctuations in fermentation parameters, the existing traceability system has limited ability to conduct in-depth analysis of brewing data and establish key parameter associations, and has failed to effectively solve the problems of taste stability and product consistency.

[0007] Therefore, how to provide a method for designing a formula for brewing tangerine red wine based on a dynamic optimization algorithm is an urgent problem that technicians in this field need to solve. Summary of the invention

[0008] One purpose of the present invention is to propose a method for designing a formula for brewing jujube red wine based on a dynamic optimization algorithm. The present invention combines a fuzzy adaptive collaborative evolution algorithm, an improved particle swarm optimization algorithm, and a traceability analysis system to achieve multi-stage intelligent optimization control of the jujube red wine brewing process. By dynamically adjusting key parameters such as temperature, humidity, and pH value at each stage, the sugar conversion rate, flavor formation, and texture balance are effectively optimized, so that the product meets high standards in terms of flavor consistency and quality stability. At the same time, the traceability analysis system tracks changes in brewing data in real time, providing scientific support for adaptive feedback optimization of subsequent batches, thereby greatly improving production efficiency and significantly improving the quality consistency and controllability of different batches of products.

[0009] A method for designing a brewing formula of tangerine red wine based on a dynamic optimization algorithm according to an embodiment of the present invention comprises the following steps:

[0010] S1, collecting the raw material composition data and brewing environment data of Citrus aurantium, and constructing the initial multidimensional data set;

[0011] S2, processing the initial multidimensional data set through the fuzzy adaptive co-evolution algorithm, analyzing the relationship between multiple parameters by combining fuzzy logic control and co-evolution mechanism, and generating the initial recipe setting;

[0012] S3, dividing the brewing process into three stages, the three stages specifically including an initial fermentation stage, a mid-term fermentation stage and a late maturation stage; the initial fermentation stage optimizes the sugar conversion rate and the initial flavor formation parameters; the mid-term fermentation stage controls the key parameters affecting the flavor generation; the late maturation stage optimizes the wine texture and flavor balance;

[0013] S4. Use the improved particle swarm optimization algorithm to control temperature, humidity, pH value and fermentation time, introduce adaptive inertia weights to control the exploration and development balance of each particle, and use a multi-agent collaborative mechanism to optimize multiple objectives in real time;

[0014] S5. Apply an adaptive variation distribution mechanism at each stage, dynamically adjust the variation range according to the optimization process, make large parameter variations in the initial fermentation stage, and gradually reduce the variation range in the later maturation stage;

[0015] S6. Optimize specific brewing goals at different brewing stages through a phased multi-objective optimization system, set the weights in the multi-objective decision-making system to be dynamically adjustable, adjust the weights of each goal in real time according to the requirements of the brewing stage, and optimize the brewing results that meet the expectations;

[0016] S7. Use the traceability analysis system to analyze the brewing data of each batch, establish the relationship between brewing parameters and finished product flavor, sweet-sour ratio and texture, and discover and track the key factors affecting product consistency;

[0017] S8. Adjust the initial formula setting for the next batch based on the traceability analysis results, and dynamically iterate the optimization model through an adaptive feedback mechanism to achieve continuous optimization of the formula.

[0018] Optionally, the S2 specifically includes:

[0019] S21, normalizing the raw material composition and brewing environment data of Citrus aurantium in the initial multidimensional data set;

[0020] S22, defining the key parameters of Huajuhong brewing as input variables in the fuzzy logic system, creating membership functions respectively and defining the range of the membership functions;

[0021] S23, constructing a fuzzy rule base, defining the relationships between variables, and generating optimized initial brewing conditions;

[0022] S24. Define multiple agents in the collaborative evolution mechanism, assign different agents to be responsible for the control of various parameters, and each agent performs independent optimization tasks based on the fuzzy rule base of the fuzzy logic control system. When each agent is initialized, it randomly generates initial values ​​related to the control variables and starts collaborative optimization:

[0023]

[0024] in, represents the parameter value of the ith agent in the t+1th generation, represents the parameter value of the ith agent in the tth generation, represents the parameter value of the jth agent in the tth generation, α represents the adaptive learning factor, and P i represents the optimal parameters of the ith agent, P j represents the optimal parameters of the jth agent, β represents the synergistic influence factor, and γ ij represents the influence coefficient between the ith agent and the jth agent, η represents the cooperative disturbance coefficient, d i represents the average distance between the ith agent and other agents, D max represents the maximum distance, and N represents the total number of agents;

[0025] S25. Evaluate the optimization effect of each agent and select the best adaptive parameters:

[0026]

[0027] Among them, F represents the fitness value, w 1 、w 2 and w 3r represents the weight coefficient, R represents the basic value of sugar conversion rate, P represents the flavor formation parameter, Q r Represents other auxiliary optimization indicators, R k The distributed sub-term representing the sugar conversion rate, θ k represents the weight of the distributed sub-item, μ represents the adjustment factor, K represents the number of distribution items of sugar conversion rate, and R represents the total number of auxiliary optimization indicators;

[0028] S26. Based on the co-evolution mechanism, the agents perform crossover and mutation operations, exchange parameter information between agents through crossover, and adjust specific parameter values ​​during the mutation process. The mutation probability is adaptively adjusted according to the optimization progress.

[0029] S27. When all agents reach a preset fitness threshold or have gone through the maximum number of evolutionary generations, the parameter combination with the highest fitness is extracted from the optimization results to generate an initial recipe setting.

[0030] Optionally, the S4 specifically includes:

[0031] S41. Define each particle as a brewing parameter combination. Each particle has an initial position and speed. Set the initial position and speed Where X i represents the parameter combination of the ith particle, V i represents the adjustment speed of the ith particle, and each particle is initialized to a randomly distributed parameter set;

[0032] S42. Establish a fitness function to evaluate the parameter combination of each particle. The fitness function is designed based on the multiple requirements of brewing and balances the effects of sugar conversion rate, temperature, humidity, pH value and fermentation time:

[0033]

[0034] Among them, A i represents the fitness of the i-th particle, ν 1 , ν 2 , ν 3 and ν 4 represents the weight coefficient, T i represents the temperature of the ith particle, H i represents the humidity of the ith particle, pH i represents the pH value of the ith particle, t i represents the fermentation parameter of the i-th particle, θ m represents the weight of the optimization target, μ represents the adjustment factor, R i,m represents the conversion rate of the i-th particle on the m-th optimization target, and M represents the total number of particles;

[0035] S43. Use adaptive inertia weight ω to control the speed update of each particle and dynamically adjust the balance between exploration and development:

[0036]

[0037] Among them, ω max represents the maximum value of the inertia weight, ω min represents the minimum value of the inertia weight, λ represents the adjustment factor, gen represents the current iteration number, max_gen represents the maximum iteration number, and ξ represents the weight perturbation factor;

[0038] S44. Update particle velocity and position based on adaptive inertia weight ω:

[0039]

[0040] in, represents the updated velocity of the ith particle, represents the velocity of the ith particle before updating, c 1 and c 2 represents the learning factor, r 1 and r 2 represents a random number between [0,1], P best,i represents the historical optimal position of the i-th particle, G best represents the global optimal position, represents the updated position of the i-th particle, Indicates the position of i particles before updating;

[0041] S45. Under the multi-agent collaborative mechanism, particles share parameter update information and adjust parameter combinations based on group feedback;

[0042] S46, repeating steps S43-S45 until the fitness values ​​of all particles reach a preset threshold, or the number of iterations reaches a maximum generation, and selecting the parameter combination with the highest fitness value.

[0043] Optionally, the S5 specifically includes:

[0044] S51. In the initial fermentation stage, a larger initial variation range is set to fully explore the multidimensional parameter space, obtain potential optimal brewing conditions, and promote the improvement of sugar conversion rate and the formation of initial flavor;

[0045] S52. In the optimization process of each generation, according to the change of fitness value, the variation range is adjusted in real time through the adaptive feedback mechanism; if the fitness is significantly improved, indicating that the current parameter combination is close to the optimal solution, the variation range is appropriately reduced to enhance the local search capability; if the fitness is limited, the variation range is appropriately increased to expand the parameter exploration range;

[0046] S53, adjusting the variation mode according to the target differences in different fermentation stages. In the early fermentation stage, a concentrated distribution variation mode is adopted to allow the parameters to vary in a larger range; in the middle fermentation stage, it is gradually switched to a fine distribution variation mode to reduce the large fluctuations of the parameters; in the late maturation stage, the variation range is further reduced;

[0047] S54. Apply diversified mutation strategies. In the early fermentation stage, adopt a random uniform distribution mutation strategy to make each parameter have an equal mutation probability and obtain a wider range of parameter combinations; in the mid-fermentation stage, introduce a group mutation strategy to group the parameters and use different mutation amplitudes for different groups to improve the optimization effect of parameter combination; in the late maturation stage, adopt a neighborhood search strategy to make parameters only slightly mutate near the current optimal value;

[0048] S55, gradually attenuate the variation range in each stage, and gradually reduce the variation range by exponential attenuation or piecewise linear attenuation according to the real-time data feedback of the brewing process. Maintain a high variation range in the initial fermentation stage, stabilize the variation range at a medium level in the middle fermentation stage, and significantly reduce the variation range in the late maturation stage as the number of iterations increases;

[0049] S56. Introduce a feedback-driven multi-stage variation range optimization mechanism to dynamically adjust the variation range by real-time monitoring of changes in fitness values ​​and improvements in brewing quality. If quality indicators improve significantly, gradually reduce the variation range and enter the fine optimization stage. If quality indicators improve slowly, increase the variation range appropriately and continue to explore better parameter combinations.

[0050] Optionally, the S6 specifically includes:

[0051] S61. In the initial fermentation stage, sugar conversion rate and initial flavor formation are set as the main optimization targets, and a higher weight value is set through the dynamic weight allocation mechanism. 1 and e 2 Prioritize controlling sugar conversion and flavor development parameters;

[0052] S62. Adjust the weight distribution of the initial fermentation stage according to the real-time data feedback. If the fitness improvement rate exceeds the preset threshold, reduce the weight distribution of the sugar conversion rate and appropriately increase the weight of the flavor parameter:

[0053]

[0054] Among them, e new represents the updated weight, e current represents the current weight, and τ represent weight adjustment factors, ΔF represents the fitness improvement rate, ΔT represents the change in sugar conversion, and F max represents the fitness target value, ζ and κ represent adjustment parameters, δ represents the attenuation coefficient, ∈ represents the smoothing term, T goal Indicates the target value of sugar conversion, T current Indicates the current sugar conversion rate;

[0055] S63. In the mid-term fermentation stage, the weights of temperature, humidity and pH are set to be dynamically adjustable, and the weights are adjusted in combination with feedback data. The adjustment of the weight coefficient follows the real-time change of the target contribution, and the target with higher fitness obtains a higher weight;

[0056] S64. Set texture and flavor balance as key goals for the late maturation stage, and set weights e based on the contribution values ​​of texture and balance in the fitness function. 3 and e 4If the texture or flavor balance is close to the expected value, the weight of the parameter is reduced and the weight is allocated to other secondary parameters to be optimized:

[0057]

[0058] Among them, e new represents the updated weight, e initial represents the initial weight, Q target represents the balance target parameter, Q current Indicates the current balance parameter, Q baseline Indicates the preset reference value, ζ e represents the deviation attenuation factor, λ e and e represents the dynamic adjustment parameter, ∈ e represents the smoothing term;

[0059] S65. Introduce an adaptive weight decay mechanism to gradually reduce the weight change amplitude of the main target in the later stage of each stage. The design of the adaptive weight decay mechanism is based on historical feedback data and the optimization process of the current stage. When the weight reaches the specified optimization threshold or the fitness is close to stability, the weight adjustment frequency is automatically reduced;

[0060] S66. According to the requirements of multi-objective optimization, the weight distribution of each target is adjusted in real time based on the feedback data of different brewing stages. When it is detected that the improvement rate of a certain target significantly exceeds that of other targets, the weight of other targets is automatically increased and the weight of the current dominant target is reduced.

[0061] Optionally, the S7 specifically includes:

[0062] S71. Collect and record key parameter data during the brewing process, including temperature T, humidity H, pH value, sugar conversion rate S and fermentation time t, and compare them with the flavor and sweetness / sourness ratio R of the finished product. sweet / sour and texture Q texture Perform associated storage;

[0063] S72. Use correlation analysis to quantify the relationship between brewing parameters and finished product characteristics, identify nonlinear associations between parameters, and use partial correlation coefficients for each pair of parameters:

[0064]

[0065] Among them, ρ T,Rsweet / sour represents the partial correlation coefficient between temperature and sweetness / sourness ratio, T i represents the temperature of the i-th batch, R sweet / sour,i represents the sweet-sour ratio of the i-th batch, T represents the mean temperature, R sweet / sour represents the mean value of sweetness-sourness ratio, α p , βp and γ p represents the balance coefficient;

[0066] S73. Establish a multiple regression model of brewing parameters and finished product characteristics to quantify the effect of each parameter on the sweet-sour ratio:

[0067] S74, using principal component analysis to perform dimensionality reduction processing on brewing data, extracting the main components that affect the flavor, sweet-sour ratio and texture of the finished product, performing cluster analysis on each batch based on the component contribution rate of the principal component analysis, using the cluster center point as a representative of different brewing conditions, and generating a cluster model in the brewing parameter space;

[0068] S75. Identify parameter fluctuations in the data sequence through anomaly detection algorithms, apply time series decomposition models to separate sudden deviations in the brewing process, trace the sudden deviations to flavor, sweet-sour ratio, and texture, establish anomaly detection factors, and record and track the source and impact of parameter deviations when a parameter fluctuation exceeds a set threshold.

[0069] The beneficial effects of the present invention are:

[0070] First, the present invention applies the fuzzy adaptive co-evolution algorithm to the brewing process, analyzes the complex relationship between brewing parameters through fuzzy logic control and co-evolution mechanism, and can perform real-time dynamic optimization of core goals such as sugar conversion, flavor generation and texture balance in different stages of fermentation, so that the brewing process can adaptively adjust key parameters according to the different needs of each stage. Especially in the early stage of brewing, the present invention improves the sugar conversion rate and promotes the formation of initial flavor; in the middle stage, it accurately controls the ratio of temperature, humidity and pH value to optimize flavor generation; in the later stage, it ensures the balanced output of wine texture and improves the consistency of finished product quality.

[0071] Secondly, the present invention introduces adaptive inertia weight control through an improved particle swarm optimization algorithm, which not only enhances the adaptability of the system in different stages of brewing, but also balances the weights of exploration and development, so that the system has stronger search capabilities and dynamic response capabilities in the parameter space. In terms of parameter variation, the adaptive variation distribution mechanism adjusts the variation range according to the brewing needs of each stage, adopts large-scale variation in the early stage to widely explore the parameter space, and gradually reduces the variation range in the middle and late stages to ensure fine adjustment, so that the entire brewing process can achieve efficient and stable optimization on the basis of ensuring quality. This highly targeted multi-objective optimization scheme significantly improves the brewing efficiency, ensures that the finished product meets the expected standards in core indicators such as flavor, texture, and sweet and sour ratio, and improves product consistency between different batches.

[0072] In addition, the traceability analysis system of the present invention establishes a deep correlation between brewing parameters and finished product characteristics, so that the system can discover and track key factors that affect product consistency. Based on the feedback and traceability analysis of each batch of data, the present invention can achieve continuous optimization of the formula and adjust the initial formula in subsequent batches in a targeted manner, thereby improving the quality of the next batch of products. This dynamic adaptive feedback mechanism not only improves the consistency and stability of the product, but also significantly reduces the risk of substandard quality and ensures high-quality output of the product. Through data recording and in-depth analysis of the entire process, the present invention also enhances the traceability of the production process, provides strong data support for subsequent formula optimization, and makes the formula design more scientific and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 A flow chart of a method for designing a brewing formula of tangerine red wine based on a dynamic optimization algorithm proposed by the present invention;

[0075] Figure 2 This is a schematic diagram of the multi-stage optimization structure of a method for designing a brewing formula for tangerine red wine based on a dynamic optimization algorithm proposed in the present invention. DETAILED DESCRIPTION

[0076] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0077] refer to Figure 1 and Figure 2 A method for designing a brewing formula of tangerine red wine based on a dynamic optimization algorithm comprises the following steps:

[0078] S1, collecting the raw material composition data and brewing environment data of Citrus aurantium, and constructing the initial multidimensional data set;

[0079] S2, processing the initial multidimensional data set through the fuzzy adaptive co-evolution algorithm, analyzing the relationship between multiple parameters by combining fuzzy logic control and co-evolution mechanism, and generating the initial recipe setting;

[0080] S3, dividing the brewing process into three stages, the three stages specifically including an initial fermentation stage, a mid-term fermentation stage and a late maturation stage; the initial fermentation stage optimizes the sugar conversion rate and the initial flavor formation parameters; the mid-term fermentation stage controls the key parameters affecting the flavor generation; the late maturation stage optimizes the wine texture and flavor balance;

[0081] S4. Use the improved particle swarm optimization algorithm to control temperature, humidity, pH value and fermentation time, introduce adaptive inertia weights to control the exploration and development balance of each particle, and use a multi-agent collaborative mechanism to optimize multiple objectives in real time;

[0082] S5. Apply an adaptive variation distribution mechanism at each stage, dynamically adjust the variation range according to the optimization process, make large parameter variations in the initial fermentation stage, and gradually reduce the variation range in the later maturation stage;

[0083] S6. Optimize specific brewing goals at different brewing stages through a phased multi-objective optimization system, set the weights in the multi-objective decision-making system to be dynamically adjustable, adjust the weights of each goal in real time according to the requirements of the brewing stage, and optimize the brewing results that meet the expectations;

[0084] S7. Use the traceability analysis system to analyze the brewing data of each batch, establish the relationship between brewing parameters and finished product flavor, sweet-sour ratio and texture, and discover and track the key factors affecting product consistency;

[0085] S8. Adjust the initial formula setting for the next batch based on the traceability analysis results, and dynamically iterate the optimization model through an adaptive feedback mechanism to achieve continuous optimization of the formula.

[0086] In this implementation, S2 specifically includes:

[0087] S21, normalizing the raw material composition and brewing environment data of Citrus aurantium in the initial multidimensional data set;

[0088] S22, defining the key parameters of Huajuhong brewing as input variables in the fuzzy logic system, creating membership functions respectively and defining the range of the membership functions;

[0089] S23, constructing a fuzzy rule base, defining the relationships between variables, and generating optimized initial brewing conditions;

[0090] S24. Define multiple agents in the collaborative evolution mechanism, assign different agents to be responsible for the control of various parameters, and each agent performs independent optimization tasks based on the fuzzy rule base of the fuzzy logic control system. When each agent is initialized, it randomly generates initial values ​​related to the control variables and starts collaborative optimization:

[0091]

[0092] in, represents the parameter value of the ith agent in the t+1th generation, represents the parameter value of the ith agent in the tth generation, represents the parameter value of the jth agent in the tth generation, α represents the adaptive learning factor, and P i represents the optimal parameters of the ith agent, P j represents the optimal parameters of the jth agent, β represents the synergistic influence factor, and γ ij represents the influence coefficient between the ith agent and the jth agent, η represents the cooperative disturbance coefficient, d i represents the average distance between the ith agent and other agents, D max represents the maximum distance, and N represents the total number of agents;

[0093] S25. Evaluate the optimization effect of each agent and select the best adaptive parameters:

[0094]

[0095] Among them, F represents the fitness value, w 1 、w 2 and w 3r represents the weight coefficient, R represents the basic value of sugar conversion rate, P represents the flavor formation parameter, Q r Represents other auxiliary optimization indicators, R k The distributed sub-term representing the sugar conversion rate, θ k represents the weight of the distributed sub-item, μ represents the adjustment factor, K represents the number of distribution items of sugar conversion rate, and R represents the total number of auxiliary optimization indicators;

[0096] S26. Based on the co-evolution mechanism, the agents perform crossover and mutation operations, exchange parameter information between agents through crossover, and adjust specific parameter values ​​during the mutation process. The mutation probability is adaptively adjusted according to the optimization progress.

[0097] S27. When all agents reach a preset fitness threshold or have gone through the maximum number of evolutionary generations, the parameter combination with the highest fitness is extracted from the optimization results to generate an initial recipe setting.

[0098] In this implementation, S4 specifically includes:

[0099] S41. Define each particle as a brewing parameter combination. Each particle has an initial position and speed. Set the initial position and speed Where X i represents the parameter combination of the ith particle, V i represents the adjustment speed of the ith particle, and each particle is initialized to a randomly distributed parameter set;

[0100] S42. Establish a fitness function to evaluate the parameter combination of each particle. The fitness function is designed based on the multiple requirements of brewing and balances the effects of sugar conversion rate, temperature, humidity, pH value and fermentation time:

[0101]

[0102] Among them, A i represents the fitness of the i-th particle, ν 1 , ν 2 , ν 3 and ν 4 represents the weight coefficient, T i represents the temperature of the ith particle, H i represents the humidity of the ith particle, pH i represents the pH value of the ith particle, t i represents the fermentation parameter of the i-th particle, θ m represents the weight of the optimization target, μ represents the adjustment factor, R i,m represents the conversion rate of the i-th particle on the m-th optimization target, and M represents the total number of particles;

[0103] S43. Use adaptive inertia weight ω to control the speed update of each particle and dynamically adjust the balance between exploration and development:

[0104]

[0105] Among them, ω max represents the maximum value of the inertia weight, ω min represents the minimum value of the inertia weight, λ represents the adjustment factor, gen represents the current iteration number, max_gen represents the maximum iteration number, and ξ represents the weight perturbation factor;

[0106] S44. Update particle velocity and position based on adaptive inertia weight ω:

[0107]

[0108] in, represents the updated velocity of the ith particle, represents the velocity of the ith particle before updating, c 1 and c 2 represents the learning factor, r 1 and r 2 represents a random number between [0,1], P best,i represents the historical optimal position of the i-th particle, G best represents the global optimal position, represents the updated position of the i-th particle, Indicates the position of i particles before updating;

[0109] S45. Under the multi-agent collaborative mechanism, particles share parameter update information and adjust parameter combinations based on group feedback;

[0110] S46, repeating steps S43-S45 until the fitness values ​​of all particles reach a preset threshold, or the number of iterations reaches a maximum generation, and selecting the parameter combination with the highest fitness value.

[0111] In this implementation manner, S5 specifically includes:

[0112] S51. In the initial fermentation stage, a larger initial variation range is set to fully explore the multidimensional parameter space, obtain potential optimal brewing conditions, and promote the improvement of sugar conversion rate and the formation of initial flavor;

[0113] S52. In the optimization process of each generation, according to the change of fitness value, the variation range is adjusted in real time through the adaptive feedback mechanism; if the fitness is significantly improved, indicating that the current parameter combination is close to the optimal solution, the variation range is appropriately reduced to enhance the local search capability; if the fitness is limited, the variation range is appropriately increased to expand the parameter exploration range;

[0114] S53, adjusting the variation mode according to the target differences in different fermentation stages. In the early fermentation stage, a concentrated distribution variation mode is adopted to allow the parameters to vary in a larger range; in the middle fermentation stage, it is gradually switched to a fine distribution variation mode to reduce the large fluctuations of the parameters; in the late maturation stage, the variation range is further reduced;

[0115] S54. Apply diversified mutation strategies. In the early fermentation stage, adopt a random uniform distribution mutation strategy to make each parameter have an equal mutation probability and obtain a wider range of parameter combinations; in the mid-fermentation stage, introduce a group mutation strategy to group the parameters and use different mutation amplitudes for different groups to improve the optimization effect of parameter combination; in the late maturation stage, adopt a neighborhood search strategy to make parameters only slightly mutate near the current optimal value;

[0116] S55, gradually attenuate the variation range in each stage, and gradually reduce the variation range by exponential attenuation or piecewise linear attenuation according to the real-time data feedback of the brewing process. Maintain a high variation range in the initial fermentation stage, stabilize the variation range at a medium level in the middle fermentation stage, and significantly reduce the variation range in the late maturation stage as the number of iterations increases;

[0117] S56. Introduce a feedback-driven multi-stage variation range optimization mechanism to dynamically adjust the variation range by real-time monitoring of changes in fitness values ​​and improvements in brewing quality. If quality indicators improve significantly, gradually reduce the variation range and enter the fine optimization stage. If quality indicators improve slowly, increase the variation range appropriately and continue to explore better parameter combinations.

[0118] In this implementation manner, S6 specifically includes:

[0119] S61. In the initial fermentation stage, sugar conversion rate and initial flavor formation are set as the main optimization targets, and a higher weight value is set through the dynamic weight allocation mechanism. 1 and e 2 Prioritize controlling sugar conversion and flavor development parameters;

[0120] S62. Adjust the weight distribution of the initial fermentation stage according to the real-time data feedback. If the fitness improvement rate exceeds the preset threshold, reduce the weight distribution of the sugar conversion rate and appropriately increase the weight of the flavor parameter:

[0121]

[0122] Among them, e new represents the updated weight, e current represents the current weight, and τ represent weight adjustment factors, ΔF represents the fitness improvement rate, ΔT represents the change in sugar conversion, and F max represents the fitness target value, ζ and κ represent adjustment parameters, δ represents the attenuation coefficient, ∈ represents the smoothing term, T goal Indicates the target value of sugar conversion, T current Indicates the current sugar conversion rate;

[0123] S63. In the mid-term fermentation stage, the weights of temperature, humidity and pH are set to be dynamically adjustable, and the weights are adjusted in combination with feedback data. The adjustment of the weight coefficient follows the real-time change of the target contribution, and the target with higher fitness obtains a higher weight;

[0124] S64. Set texture and flavor balance as key goals for the late maturation stage, and set weights e based on the contribution values ​​of texture and balance in the fitness function. 3 and e 4 If the texture or flavor balance is close to the expected value, the weight of the parameter is reduced and the weight is allocated to other secondary parameters to be optimized:

[0125]

[0126] Among them, e new represents the updated weight, einitial represents the initial weight, Q target represents the balance target parameter, Q current Indicates the current balance parameter, Q baseline Indicates the preset reference value, ζ e represents the deviation attenuation factor, λ e and e represents the dynamic adjustment parameter, ∈ e represents the smoothing term;

[0127] S65. Introduce an adaptive weight decay mechanism to gradually reduce the weight change amplitude of the main target in the later stage of each stage. The design of the adaptive weight decay mechanism is based on historical feedback data and the optimization process of the current stage. When the weight reaches the specified optimization threshold or the fitness is close to stability, the weight adjustment frequency is automatically reduced;

[0128] S66. According to the requirements of multi-objective optimization, the weight distribution of each target is adjusted in real time based on the feedback data of different brewing stages. When it is detected that the improvement rate of a certain target significantly exceeds that of other targets, the weight of other targets is automatically increased and the weight of the current dominant target is reduced.

[0129] In this implementation manner, the S7 specifically includes:

[0130] S71. Collect and record key parameter data during the brewing process, including temperature T, humidity H, pH value, sugar conversion rate S and fermentation time t, and compare them with the flavor and sweetness / sourness ratio R of the finished product. sweet / sour and texture Q texture Perform associated storage;

[0131] S72. Use correlation analysis to quantify the relationship between brewing parameters and finished product characteristics, identify nonlinear associations between parameters, and use partial correlation coefficients for each pair of parameters:

[0132]

[0133] Among them, ρ T,Rsweet / sour represents the partial correlation coefficient between temperature and sweetness / sourness ratio, T i represents the temperature of the i-th batch, R sweet / sour,i represents the sweet-sour ratio of the i-th batch, represents the mean temperature, represents the mean value of sweetness-sourness ratio, α p , β p and γ p represents the balance coefficient;

[0134] S73. Establish a multiple regression model of brewing parameters and finished product characteristics to quantify the effect of each parameter on the sweet-sour ratio:

[0135] S74, using principal component analysis to perform dimensionality reduction processing on brewing data, extracting the main components that affect the flavor, sweet-sour ratio and texture of the finished product, performing cluster analysis on each batch based on the component contribution rate of the principal component analysis, using the cluster center point as a representative of different brewing conditions, and generating a cluster model in the brewing parameter space;

[0136] S75. Identify parameter fluctuations in the data sequence through anomaly detection algorithms, apply time series decomposition models to separate sudden deviations in the brewing process, trace the sudden deviations to flavor, sweet-sour ratio, and texture, establish anomaly detection factors, and record and track the source and impact of parameter deviations when a parameter fluctuation exceeds a set threshold.

[0137] Embodiment 1:

[0138] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a winery with large temperature changes throughout the year. The orange red wine produced by the factory requires a mellow flavor, appropriate sour and sweet, balanced texture, and must meet certain physical and chemical indicators. However, in production, due to the differences in raw material ingredients in different batches, fluctuations in environmental humidity and temperature, and uncontrollable factors in the brewing process, the sour-sweet ratio of the wine is often unstable and the flavor is quite different. Especially in the summer when the temperature is high, the difficulty of temperature control in the brewing process increases, which further affects the sugar conversion rate and the final taste consistency.

[0139] To solve these problems, the present invention is applied in the brewing process of the Huajuhong winery, and the specific operating steps are as follows. First, in the data collection stage, the brewing environment is monitored by a real-time sensor network, and the Huajuhong raw material composition and fermentation environment data, including temperature, humidity, pH value and sugar content, are continuously recorded. The collected data is standardized and an initial multidimensional data set is established. Secondly, a fuzzy adaptive co-evolutionary algorithm is used to analyze the complex relationship between the brewing parameters through fuzzy logic control and co-evolutionary mechanism, and the optimized initial formula setting is generated to ensure the optimal setting value of sugar conversion rate, flavor generation and texture balance. Subsequently, the brewing process is divided into three stages: initial fermentation, mid-term fermentation and late maturation, and each stage has different control focuses. The initial stage focuses on sugar conversion rate and initial flavor formation, the mid-term focuses on fine control of temperature and humidity to enhance the flavor generation effect, and the late stage focuses on texture balance to ensure the consistency of the taste of the finished product.

[0140] In actual operation, an improved particle swarm optimization algorithm is applied to control brewing temperature, humidity and pH value in real time. Each parameter is independently controlled by the agent, which continuously adjusts the set value based on real-time feedback. In the early fermentation stage, the exploration weight of the particle swarm is increased to ensure sufficient sugar conversion. In the mid-fermentation stage, the agent adjusts the parameters according to the temperature, humidity and pH value feedback from the sensor, avoiding the influence of high temperature on flavor generation and achieving an ideal sour-sweet ratio.

[0141] To further verify the effectiveness of the present invention, the traceability analysis system was applied to brewing data from different batches. By analyzing the brewing parameters of each batch and the flavor characteristics of the finished product, the traceability system established a correlation model between key indicators such as sugar conversion rate, sweet-sour ratio and texture and brewing parameters. Under this feedback control, the flavor stability of the finished product was significantly improved, while reducing the cost and time of readjustment.

[0142] Table 1 Comparison of experimental data of traditional process and the method of the present invention in the process of brewing orange red wine

[0143]

[0144]

[0145] According to Table 1 above, we can clearly see the significant improvements in different brewing parameters of the method of the present invention, especially the improvement in sugar conversion rate, sweet-sour ratio, texture consistency index and brewing cycle under different seasonal conditions. First, in terms of sugar conversion rate, the conversion rates of the traditional process in summer and winter were 72.5% and 68.2% respectively, while after using the method of the present invention, the sugar conversion rate in summer was increased to 83.4%, and in winter it reached 80.6%, indicating that the optimized multi-stage dynamic control has achieved a significant improvement in conversion rate, especially in summer when the temperature fluctuates greatly, the effect is more obvious.

[0146] The sweet-sour ratio is an important parameter for measuring the flavor of citrus red wine. When using traditional technology, the sweet-sour ratio in summer and winter fluctuates between 1.75 and 1.6, respectively. The method of the present invention optimizes it to 1.5 and 1.45 respectively through refined control, and the fluctuation range is significantly reduced. This improvement in consistency shows that the improved particle swarm optimization algorithm and multi-objective dynamic weight adjustment mechanism have a good regulatory effect on flavor control.

[0147] Texture consistency is an important manifestation of the quality stability of Huajuhong wine. Under the traditional process, the texture consistency indexes in summer and winter were 6.3 and 6.1 respectively, while under the optimized control of the method of the present invention, they were increased to 8.7 and 8.9 respectively in summer and winter, showing better wine body balance and texture stability. This significant improvement reflects the control effect of the traceability analysis and adaptive feedback mechanism in the present invention in the multi-batch brewing process.

[0148] Finally, in terms of brewing cycle, the method of the present invention also shows good optimization effect. The brewing cycle of the traditional process is 50 days in summer and 45 days in winter, while the method of the present invention shortens the brewing cycle to 45 days and 40 days respectively. This improvement not only effectively improves production efficiency, but also significantly reduces production costs, reflecting the efficiency and application potential of the method of the present invention in industrial production.

[0149] The experimental data clearly show that the method of the present invention is superior to the traditional process in terms of sugar conversion rate, flavor control, texture consistency and production cycle, further verifying its practical value in achieving high-quality and standardized production of jujube red wine.

[0150] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for designing a formula for brewing tangerine red wine based on a dynamic optimization algorithm, characterized in that: The steps include: S1, collecting the raw material composition data and brewing environment data of Citrus aurantium, and constructing the initial multidimensional data set; S2, processing the initial multidimensional data set through the fuzzy adaptive co-evolution algorithm, analyzing the relationship between multiple parameters by combining fuzzy logic control and co-evolution mechanism, and generating the initial recipe setting; S3, dividing the brewing process into three stages, the three stages specifically including an initial fermentation stage, a mid-term fermentation stage and a late maturation stage; the initial fermentation stage optimizes the sugar conversion rate and the initial flavor formation parameters; the mid-term fermentation stage controls the key parameters affecting the flavor generation; the late maturation stage optimizes the wine texture and flavor balance; S4. Use the improved particle swarm optimization algorithm to control temperature, humidity, pH value and fermentation time, introduce adaptive inertia weights to control the exploration and development balance of each particle, and use a multi-agent collaborative mechanism to optimize multiple objectives in real time; S5. Apply an adaptive variation distribution mechanism at each stage, dynamically adjust the variation range according to the optimization process, make large parameter variations in the initial fermentation stage, and gradually reduce the variation range in the later maturation stage; S6. Optimize specific brewing goals at different brewing stages through a phased multi-objective optimization system, set the weights in the multi-objective decision-making system to be dynamically adjustable, adjust the weights of each goal in real time according to the requirements of the brewing stage, and optimize the brewing results that meet the expectations; S7. Use the traceability analysis system to analyze the brewing data of each batch, establish the relationship between brewing parameters and finished product flavor, sweet-sour ratio and texture, and discover and track the key factors affecting product consistency; S8. Adjust the initial formula setting for the next batch based on the traceability analysis results, and dynamically iterate the optimization model through an adaptive feedback mechanism to achieve continuous optimization of the formula.

2. A method for designing a formula for brewing orange red wine based on a dynamic optimization algorithm according to claim 1, characterized in that: The S2 specifically includes: S21, normalizing the raw material composition and brewing environment data of Citrus aurantium in the initial multidimensional data set; S22, defining the key parameters of Huajuhong brewing as input variables in the fuzzy logic system, creating membership functions respectively and defining the range of the membership functions; S23, constructing a fuzzy rule base, defining the relationships between variables, and generating optimized initial brewing conditions; S24. Define multiple agents in the collaborative evolution mechanism, assign different agents to be responsible for the control of various parameters, and each agent performs independent optimization tasks based on the fuzzy rule base of the fuzzy logic control system. When each agent is initialized, it randomly generates initial values ​​related to the control variables and starts collaborative optimization: in, represents the parameter value of the ith agent in the t+1th generation, represents the parameter value of the ith agent in the tth generation, represents the parameter value of the jth agent in the tth generation, α represents the adaptive learning factor, and P i represents the optimal parameters of the ith agent, P j represents the optimal parameters of the jth agent, β represents the synergistic influence factor, and γ ij represents the influence coefficient between the ith agent and the jth agent, η represents the cooperative disturbance coefficient, d i represents the average distance between the ith agent and other agents, D max represents the maximum distance, and N represents the total number of agents; S25. Evaluate the optimization effect of each agent and select the best adaptive parameters: Among them, F represents the fitness value, w1, w2 and w 3r represents the weight coefficient, R represents the basic value of sugar conversion rate, P represents the flavor formation parameter, Q r Represents other auxiliary optimization indicators, R k The distributed sub-term representing the sugar conversion rate, θ k represents the weight of the distributed sub-item, μ represents the adjustment factor, K represents the number of distribution items of sugar conversion rate, and R represents the total number of auxiliary optimization indicators; S26. Based on the co-evolution mechanism, the agents perform crossover and mutation operations, exchange parameter information between agents through crossover, and adjust specific parameter values ​​during the mutation process. The mutation probability is adaptively adjusted according to the optimization progress. S27. When all agents reach a preset fitness threshold or have gone through the maximum number of evolutionary generations, the parameter combination with the highest fitness is extracted from the optimization results to generate an initial recipe setting.

3. A method for designing a formula for brewing orange red wine based on a dynamic optimization algorithm according to claim 1, characterized in that: The S4 specifically includes: S41. Define each particle as a brewing parameter combination. Each particle has an initial position and speed. Set the initial position and speed Where X i represents the parameter combination of the ith particle, V i represents the adjustment speed of the ith particle, and each particle is initialized to a randomly distributed parameter set; S42. Establish a fitness function to evaluate the parameter combination of each particle. The fitness function is designed based on the multiple requirements of brewing and balances the effects of sugar conversion rate, temperature, humidity, pH value and fermentation time: Among them, A i represents the fitness of the i-th particle, ν1, ν2, ν3 and ν4 represent weight coefficients, T i represents the temperature of the ith particle, H i represents the humidity of the ith particle, pH i represents the pH value of the ith particle, t i represents the fermentation parameter of the i-th particle, θ m represents the weight of the optimization target, μ represents the adjustment factor, R i,m represents the conversion rate of the i-th particle on the m-th optimization target, and M represents the total number of particles; S43. Use adaptive inertia weight ω to control the speed update of each particle and dynamically adjust the balance between exploration and development: Among them, ω max represents the maximum value of the inertia weight, ω min represents the minimum value of the inertia weight, λ represents the adjustment factor, gen represents the current iteration number, max_gen represents the maximum iteration number, and ξ represents the weight perturbation factor; S44. Update particle velocity and position based on adaptive inertia weight ω: in, represents the updated velocity of the ith particle, represents the speed of the i-th particle before updating, c1 and c2 represent learning factors, r1 and r2 represent random numbers between [0,1], P best,i represents the historical optimal position of the i-th particle, G best represents the global optimal position, represents the updated position of the i-th particle, Indicates the position of i particles before updating; S45. Under the multi-agent collaborative mechanism, particles share parameter update information and adjust parameter combinations based on group feedback; S46, repeating steps S43-S45 until the fitness values ​​of all particles reach a preset threshold, or the number of iterations reaches a maximum generation, and selecting the parameter combination with the highest fitness value.

4. The method for designing a formula for brewing orange red wine based on a dynamic optimization algorithm according to claim 1, characterized in that: The S5 specifically includes: S51. In the initial fermentation stage, a larger initial variation range is set to fully explore the multidimensional parameter space, obtain potential optimal brewing conditions, and promote the improvement of sugar conversion rate and the formation of initial flavor; S52. In the optimization process of each generation, according to the change of fitness value, the variation range is adjusted in real time through the adaptive feedback mechanism; if the fitness is significantly improved, indicating that the current parameter combination is close to the optimal solution, the variation range is appropriately reduced to enhance the local search capability; if the fitness is limited, the variation range is appropriately increased to expand the parameter exploration range; S53, adjusting the variation mode according to the target differences in different fermentation stages. In the early fermentation stage, a concentrated distribution variation mode is adopted to allow the parameters to vary in a larger range; in the middle fermentation stage, it is gradually switched to a fine distribution variation mode to reduce the large fluctuations of the parameters; in the late maturation stage, the variation range is further reduced; S54. Apply diversified mutation strategies. In the early fermentation stage, adopt a random uniform distribution mutation strategy to make each parameter have an equal mutation probability and obtain a wider range of parameter combinations; in the mid-fermentation stage, introduce a group mutation strategy to group the parameters and use different mutation amplitudes for different groups to improve the optimization effect of parameter combination; in the late maturation stage, adopt a neighborhood search strategy to make parameters only slightly mutate near the current optimal value; S55, gradually attenuate the variation range in each stage, and gradually reduce the variation range by exponential attenuation or piecewise linear attenuation according to the real-time data feedback of the brewing process. Maintain a high variation range in the initial fermentation stage, stabilize the variation range at a medium level in the middle fermentation stage, and significantly reduce the variation range in the late maturation stage as the number of iterations increases; S56. Introduce a feedback-driven multi-stage variation range optimization mechanism to dynamically adjust the variation range by real-time monitoring of changes in fitness values ​​and improvements in brewing quality. If quality indicators improve significantly, gradually reduce the variation range and enter the fine optimization stage. If quality indicators improve slowly, increase the variation range appropriately and continue to explore better parameter combinations.

5. The method for designing a formula for brewing orange red wine based on a dynamic optimization algorithm according to claim 1, characterized in that: The S6 specifically includes: S61. In the initial fermentation stage, sugar conversion rate and initial flavor formation are set as the main optimization targets, and higher weight values ​​e1 and e2 are set through a dynamic weight allocation mechanism to control the priority of sugar conversion rate and flavor formation parameters; S62. Adjust the weight distribution of the initial fermentation stage according to the real-time data feedback. If the fitness improvement rate exceeds the preset threshold, reduce the weight distribution of the sugar conversion rate and appropriately increase the weight of the flavor parameter: Among them, e new represents the updated weight, e current represents the current weight, and τ represent weight adjustment factors, ΔF represents the fitness improvement rate, ΔT represents the change in sugar conversion, and F max represents the fitness target value, ζ and κ represent adjustment parameters, δ represents the attenuation coefficient, ∈ represents the smoothing term, T goal Indicates the target value of sugar conversion, T current Indicates the current sugar conversion rate; S63. In the mid-term fermentation stage, the weights of temperature, humidity and pH are set to be dynamically adjustable, and the weights are adjusted in combination with feedback data. The adjustment of the weight coefficient follows the real-time change of the target contribution, and the target with higher fitness obtains a higher weight; S64, setting texture and flavor balance as key goals for the late maturation stage, setting priorities for weights e3 and e4 based on the texture and balance contribution values ​​in the fitness function, and if it is detected that the degree of achievement of texture or flavor balance is close to the expected value, reducing the weight of the parameter and assigning the weight to other secondary parameters to be optimized: Among them, e new represents the updated weight, e initial represents the initial weight, Q target represents the balance target parameter, Q current Indicates the current balance parameter, Q baseline Indicates the preset reference value, ζ e represents the deviation attenuation factor, λ e and e represents the dynamic adjustment parameter, ∈ e represents the smoothing term; S65. Introduce an adaptive weight decay mechanism to gradually reduce the weight change amplitude of the main target in the later stage of each stage. The design of the adaptive weight decay mechanism is based on historical feedback data and the optimization process of the current stage. When the weight reaches the specified optimization threshold or the fitness is close to stability, the weight adjustment frequency is automatically reduced; S66. According to the requirements of multi-objective optimization, the weight distribution of each target is adjusted in real time based on the feedback data of different brewing stages. When it is detected that the improvement rate of a certain target significantly exceeds that of other targets, the weight of other targets is automatically increased and the weight of the current dominant target is reduced.

6. The method for designing a formula for brewing orange red wine based on a dynamic optimization algorithm according to claim 1, characterized in that: The S7 specifically includes: S71. Collect and record key parameter data during the brewing process, including temperature T, humidity H, pH value, sugar conversion rate S and fermentation time t, and compare them with the flavor and sweetness / sourness ratio R of the finished product. sweet / sour and texture Q texture Perform associated storage; S72. Use correlation analysis to quantify the relationship between brewing parameters and finished product characteristics, identify nonlinear associations between parameters, and use partial correlation coefficients for each pair of parameters: in, represents the partial correlation coefficient between temperature and sweetness / sourness ratio, T i represents the temperature of the i-th batch, R sweet / sour,i represents the sweet-sour ratio of the i-th batch, represents the mean temperature, represents the mean value of sweetness-sourness ratio, α p , β p and γ p represents the balance coefficient; S73. Establish a multiple regression model of brewing parameters and finished product characteristics to quantify the effect of each parameter on the sweet-sour ratio: S74, using principal component analysis to perform dimensionality reduction processing on brewing data, extracting the main components that affect the flavor, sweet-sour ratio and texture of the finished product, performing cluster analysis on each batch based on the component contribution rate of the principal component analysis, using the cluster center point as a representative of different brewing conditions, and generating a cluster model in the brewing parameter space; S75. Identify parameter fluctuations in the data sequence through anomaly detection algorithms, apply time series decomposition models to separate sudden deviations in the brewing process, trace the sudden deviations to flavor, sweet-sour ratio, and texture, establish anomaly detection factors, and record and track the source and impact of parameter deviations when a parameter fluctuation exceeds a set threshold.

Images